Concurrent with the increase in sophistication and drive towards miniaturization of microelectronics, the number of transistors per integrated circuit has exponentially grown and promises to grow to meet the demands for faster, smaller and more powerful electronic systems. However, as traditional silicon-based transistor geometries reach a critical point where the silicon dioxide gate dielectric becomes just a few atomic layers thick, tunneling of electrons will become more prevalent leading to current leakage and increase in power dissipation. Accordingly, an alternative dielectric possessing a higher permittivity or dielectric constant than silicon dioxide and capable of preventing current tunneling or leakage would be highly desirable. Among the most promising dielectric candidates to replace silicon dioxide includes hafnium oxide, zirconium oxide and tantalum oxide.
Unfortunately, these materials are chemically and thermally unstable on silicon, unlike silicon dioxide, forming defects and charge traps at the interface between the metal dielectric and the silicon substrate. The charge traps and defects absorb the voltage applied at the gate and perturb the performance and reliability of the transistor. To limit the formation of interfacial charge traps and defects, an interfacial layer of silicon dioxide is deposited in between the dielectric and the silicon substrate. The silicon dioxide interface buffers the silicon substrate from the dielectric, but the silicon dioxide interface may not be compatible with the surface properties of the dielectric. Accordingly, an interface that can ameliorate the surface properties and chemistries of the dielectric and silicon substrate, while minimizing the equivalent physical oxide thickness, is needed to fabricate ultra-thin high k dielectrics.
Prior art deposition techniques for fabricating films such as chemical vapor deposition (CVD) are increasingly unable to meet the requirements of advanced thin films. While CVD processes can be tailored to provide conformal films with improved step coverage, CVD processes often require high processing temperatures. For instance, one of the obstacles of making high k gate dielectrics is the formation of an interfacial silicon oxide layer during CVD processes. Gas phase reaction in CVD leads to particle generation. Another obstacle is the limitation of prior art CVD processes in depositing ultra thin films for high k gate dielectrics on a silicon substrate.
An alternative to traditional CVD processes to deposit very thin films is Atomic layer deposition (ALD). ALD has several advantages over traditional CVD. ALD can be performed at comparatively low temperatures which is compatible with the industry's trend toward lower temperatures, and can produce conformal thin film layers. he existing method for depositing multi-component films, such as a HfxSiyO2(x+y=1) film, using an ALD processes is to deposit laminate films of HfO2 and SiO2 film using a sequential vapor deposition method. That is, the precursor chemicals are not mixed, and instead an Hf containing precursor and a Si containing precursor are pulsed independently and sequentially into the chamber to form laminate layers of HfO and SiO2, respectively. In fact, any mixing of precursors is prohibited, and the chamber is purged of one precursor before the second precursor is pulsed. Once the laminate films are formed to a desired thickness, the film is annealed in an attempt to arrive at more continuous composition throughout the film. This approach of building up layers of different laminate films leads to many electron traps in the film due to the multiple interfaces which requires high temperature thermal anneal to fix the traps. The addition of the high temperature thermal annealing step increases cost and time for manufacturing semiconductors, and moreover can result in the undesirable out migration of elements from previously formed layers on the wafer. In addition, it is difficult to control the stoichiometric composition of multi-component films in the laminate method. The dielectric constant (k), crystallization temperature and refractive index of HfSiOx films cannot be easily controlled by the traditional one chemical sequential precursor pulse methods (such as the laminate method). Furthermore, the cycle times needed to form a film of desired thickness using the conventional sequential pulse and purge of one chemical precursor at a time are impractical and require too much time for future IC manufacturing.
Attempts to fabricate a multi-component films using mixed precursors have been limited to the traditional CVD methods. For example, U.S. Pat. Nos. 6,537,613 and 6,238,734 both to Senzaki et al. (the '613 and '734 patents) generally disclose system and methods for generating a compositional gradient comprising a metal and metalloid compound by direct liquid injection. In direct liquid injection (DLI), the metal and metalloid precursors are mixed together to form a solventless liquid mixture prior to injection of the mixture into the deposition system.
There are however several drawbacks associated with the method described in the '613 and '734 patents. Specifically, it is a liquid mixture that is injected. As such, if the liquid mixture is not thoroughly mixed, a film having an uneven composition and gradient will form on the substrate. In addition, even if appropriate volumes of samples are provided, there is no guarantee that the mixture will vaporize uniformly since each precursor has a unique boiling point, vapor pressure and volatility. Furthermore, if the discrepancy in boiling points between the precursors is substantial, one precursor may decompose at the boiling point of the second forming particulates or contaminants. Generally, either the precursors have not been adequately mixed, resulting in a non-uniform film composition, or mixing of the two vapors causes pre-reaction in the gas phase, resulting in the formation of particles or contaminants that are deposited on the wafer.
Accordingly, there is a need for further developments in methods of fabricating multi-component films. There is particularly a need for a method of fabricating multi-component films using an ALD process. It is further desirable that the method provides control of the stoichiometric composition or gradient of a multi-component film.